mpls technology overview

54
RMDCN workshop on the migration to IP/VPN 1/54 MPLS Technology Overview Ahmed Benallegue [email protected]

Upload: others

Post on 19-Jan-2022

7 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 1/54

MPLS Technology Overview

Ahmed [email protected]

Page 2: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 2/54

Plan

1. MPLS basics2. The MPLS approach3. Label distribution –RSVP-TE4. Traffic Engineering5. QoS, CoS and DiffServ (DS)6. “DS-Aware”MPLS Traffic Engineering (DS-TE)7. MPLS vs. FRQuestions

Page 3: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 3/54

1. MPLS Basics

MPLA BasicsWhy MPLSWhat is MPLS

Page 4: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 4/54

1. MPLS Basics –Why MPLS

Hypothesis: all networks are or will be Internets (IP based)MPLS was defined first and foremost as an IP-centric solutionA multi-protocol solution is important

Many legacy protocols will still be used

0

10

20

30

40

50

60

70

80

1994 1996 1998 2000 2002

Per

cen

to

fT

raff

ic

IP

SNA

IPX

Other

Page 5: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 5/54

1. MPLS Basics –Why MPLS (cont)Traditional IP forwarding

Uses hop-to-hop routing/forwarding Inefficient as Layer 3 lookup required for each packetHop-by-hop networks are difficult to control

How do we guarantee QoS in a connectionless networkDifficult to control congestion in a large hop-by-hop networks

Page 6: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 6/54

1. MPLS Basics –Why MPLS (cont)

Overcoming the limitations of traditional IPNew ways of forwarding packets

Alternative to hop-by-hop L3 lookups

New ways of allocating resourcesNecessary to provide guaranteed QoS

Integrate routing and switchingSwitch = high speed forwardingConnection oriented solutions: allocation of resources per virtual

connectionRouting performed once at the Virtual Circuit (VC) setup

Solution: make L2 switches “IP aware” This is what MPLS does

Page 7: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 7/54

1. MPLS Basics –What is MPLS

MPLS evolution First used to facilitate new IP networks

QoS, scalability, traffic engineering Then MPLS was leveraged to provide IPVPNs

Or L3 MPLS VPNNow heading towards a converged core for L2 and L3 services

L2 VPNs and support of legacy, e.g.: ATM, FR, Ethernet

What is MPLS Label Switching

Data units are switched (routed) through the network by reference toan attached label

This in itself is not new: ATM and FR are label switched networksMulti-Protocol

It can handle multiple different network layer protocolIt can utilise different L2 technologies: ATM, FR, Ethernet, PPP…

Page 8: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 8/54

1. MPLS Basics –What is MPLS (cont)Label switching concept

Simplifies data forwardingRemoves the need to examine the L3 (IP) headers at each hop

User data is “labelled” The label determines the route taken through the network

Labels have local significance

Page 9: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 9/54

1. MPLS Basics –What is MPLS (cont)

Multi-Protocol Initial emphasis on IPNew focus on convergence to carry legacy services/protocols

Page 10: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 10/54

1. MPLS Basics –What is MPLS (cont)

What does MPLS achieve? Integrate IP routing and L2 switching

Can reuse existing switching techniques such as FR and ATMLabel switching

Enables traffic engineering in IP networksRoutes other than shortest path can be usedDefine explicit paths through a networkTraffic Engineering (vs. Network Engineering)

–Put traffic where there is bandwidthSupports different services and QoS

Achieving QoS requires classification, marking and resource allocationon a path.

Resource Reservation Protocol-Tunnelling Extensions (RSVP-TE)Differentiated Services (DiffServ or DS)

Page 11: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 11/54

2. MPLS Approach

The MPLS approachArchitecture Forwarding modelGeneric MPLS headerMPLS label for FR Label stack example

Page 12: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 12/54

2. MPLS Approach –Architecture

Basic MPLS network architectureBasic operation

Ingress Label Edge Router (LER) adds labels to unlabelled packetsLabel Switched Router (LSR) forwards labelled packets following an

unidirectional Label Switch Path (LSP)Path determined by routing, established by signallingEgress LER removes label and forwards packetsLSRs switch labelled packets and route IP packets

Two preconditionsThe LSRs must participate in IP routingThe LSRs must participate in Label Distribution

Page 13: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 13/54

2. MPLS Approach –Architecture (cont)

Basic MPLS network architecture (cont.)

Page 14: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 14/54

2. MPLS Approach –Forwarding model

Conventional IP forwarding• Input router receives IP packet• Lookup to find matching IP network address• Lookup to find output port• Send packet to next hop router and repeat• Packet reaches destination router• Packet delivered to destination (ARP etc.)

Page 15: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 15/54

2. MPLS Approach –Forwarding model (cont)

MPLS forwarding • LER receives IP packet and does IP routing table lookup• Assigns packet to Forwarding Equivalence Class (FEC)• Assigns FEC to a label and labels the packet• Forwards packet to LSR1• LSR1 looks up table to find OUT port and Label OUT• Process repeated at each LSR - LSRs do not look up IP address• Label stripped on exit

Page 16: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 16/54

Constructing the forwarding tableStandard routing protocol used to test the status of neighbours Label Distribution Protocols used to bind between labels and

FEC

2. MPLS Approach –Forwarding model (cont)

Page 17: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 17/54

2. MPLS Approach –Generic MPLS Header

With L2 protocol that does not have label fieldMPLS header is inserted between the L2 and IP headers

The “Exp.”(or experimental) 3-bits field used for CoSThe “Stack”(S) field is set to 1 in the last entry in the stack

Forwarding is done with reference to the top label

Page 18: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 18/54

2. MPLS Approach –MPLS Label for FR

With the FR header structure The DLCI field is used to carry the current (top) label The MPLS header is used to carry TTL information FR switches (acting as LSR) do not see the MPLS headers

Page 19: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 19/54

2. MPLS Approach –Label Stack ExampleTwo LSPs: A to C and B to D

The second level labels remain unchanged

Two main characteristics Aggregation: several LSPs into one LSP (or “tunnel”) LSRs in the core network do not have visibility outside the core

Useful for the support of VPN (two VPNs in this example: A and C, and B and D

Page 20: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 20/54

3. Label Distribution

Label distribution Two flavoursRSVP-TE

Page 21: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 21/54

3. Label Distribution –Two flavours

Hop-by-Hop LDP Label Distribution Based on “normal”IP Time taken to re-route

(depends on convergence ofrouting protocol)

LDP does not supportresource allocation (QoS) ortraffic engineering

Explicit Routing RSVP-TE Label Distribution Uses source routing Specified path from source to

destination Re-routing can be achieved

quickly Used for allocating resources

(QoS) and performing trafficengineering

Hop-by-Hop: basic MPLS network Most common protocol for Traffic Engineering LSP is RSVP-TE

Resource Reservation Protocol with Tunnel Extensions

Page 22: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 22/54

3. Label Distribution –RSVP-TE

The flow conceptDatagrams are part of a sequence from source to destinationNormal forwarding: each datagram is dealt with independentlyA flow identifies a sequence of packets

Flow requires routers to “remember”state information associated with it

RSVP-TE is a signalling protocol devised before MPLSRSVP-TE addresses the requirements to ensure QoS

Pre-allocating resources along the path to support a flowReport when resources not available

Standard routing processes are not sufficient They are reactive rather than predictive

Page 23: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 23/54

3. Label Distribution –RSVP-TE (cont.)

RSVP PATH message: Session: unique ID for the LSP Explicit Route: Specify route from

Ingress to Egress Record Route: listing of the LSRs

traversed by the LSP

RSVP RESERVATIONmessage Label: perform the label

distribution upstream Style: specify the reservation style

Fixed filter: dedicated BW Shared explicit: shared BW

Record Route: Return the path tothe Ingress LSR (A)

RSVP-TE signalling takes place between Ingress and Egress Establish an LSP tunnel

Page 24: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 24/54

3. Label Distribution –RSVP-TE (cont.)

Label Distribution and Binding1. LSR B receives RESV and binds the label to the port (30, Y)2. LSR B allocates a local label (20) and binds it to the port (X)3. LSR B generates a new LABEL object, replaces the previous

one (30) and sends to previous LSR.

Page 25: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 25/54

3. Label Distribution –RSVP-TE (cont.)

LSP re-routing1. The initial LSP has STYLE = Shared Explicit2. New Explicit Route reservation AB, BC, CD, DE and

EF3. When new LSP is established

Ingress moves the flow to new LSP Initial LSP is torn down

Page 26: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 26/54

4. Traffic Engineering

Traffic EngineeringRequirementsBasics

Page 27: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 27/54

4. Traffic Engineering –Requirements

Control traffic flows in the networkBetter utilise network resources, move traffic to specified pathsEnsure QoS requirements are met

Establish LSPs: activate and deactivate LSPsProvide re-routing capabilities

Re-routing due to failed pathsPre-emption

Provide details traffic statistics for optimisation and futurecapacity planning

Traffic engineering vs. network engineering: BOTH!Network engineering to provide BWTraffic engineering to use BW efficiently

Page 28: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 28/54

4. Traffic Engineering –Basics

Automated LSP path selectionUse RSVP-TE signalling protocols to establish constraint-based

LSPs

Constraint determined by either/both: The network operator

Required BW, include or exclude specific LSPs from the LSP

Information in a Traffic Engineering Database

Page 29: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 29/54

4. Traffic Engineering –Basics (cont.)

An Explicit Route (ER) is setup by the Ingress LSR IGP Extended for TE: max and remaining “reservabale”BW TE Database: input from the extended IGP (OSPF or IS-IS)User defined constraints: the network operator (user) can define

BWHop limitsSetup and Holding prioritiesExplicit Route definition

Shortest Path calculationLinks selection

Explicit RouteThe resulting route of SPF calculation

LSP SignallingSetup of the Explicit Route LSP

IGP Extended for Traffic Engineering

Routing Table TE Database

Shortest Path (Constrained) calculation

Explicit Route

LSP Signalling

User definedconstraints

Page 30: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 30/54

4. Traffic Engineering –Basics (cont.)

Recovery methodsReroutingProtection Switching

Common technique used by transmission systems, e.g. Optical and SDHIn MPLS: swapping traffic to a backup LSP within 50msecRelies on failure detection mechanisms

Use pre-established LSPsBackup may be between

Ingress and Egress: global repairAny LSRs: local repair

Page 31: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 31/54

5. QoS & CoS

QoS & CoSDefinitions The DiffServ (DS) modelDiffServ and ToSDiffServ basicsDiffServ and MPLS

Page 32: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 32/54

5. QoS & CoS –Definitions

QoS vs. CoSQoS provides absolute quantifiable levels of performance

E.g. based on parameters such as throughput, delay…

CoS provides for relative levels of performanceIt assigns traffic flow to classes where different classes receive

different treatment

QoS requires guaranteed BW and class-based forwarding It is useless to have separate classes if there is no BW

CoS only requires class-based forwarding

Both Traffic Engineered LSP and class-based forwardingare required to provide QoS in an MPLS environment

Page 33: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 33/54

5. QoS & CoS –The DiffServ Model

Why differentiated services? Traditional IP is best-effortAll users get the same service: no distinction between the

different trafficsNew applications require specific QoSService providers need to differentiate the different flowsCustomers want to request QoS

The DiffServ (DS) modelA simple way to differentiate servicesClassify traffic entering the networkAssign traffic to a service classMark traffic with its service class at ingress Inside the network treat packets according to their service class

Page 34: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 34/54

5. QoS & CoS –DiffServ and ToS

IP’s Type of Service (ToS) fieldCan specify relative QoS, coupled to routing In practice it has not been used consistently

The DS fieldRedefined ToS: the 6 most significant bits (64 possible values)DiffServ Code Point (DSCP) is the value of the DS fieldDSCP of 000000 is a default for best-effort service

Page 35: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 35/54

5. QoS & CoS –DiffServ Basics

The Ingress routerClassifies trafficConditions trafficPolices trafficPuts traffic in priority queues

Based on the marking

All packets are given the same treatment within the networkDiffServ basics

Moves complexity to ingress: traffic classification and markingRequires node to examine the DS fieldDS marking applied to aggregate flows

Page 36: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 36/54

5. QoS & CoS –DiffServ and MPLS

Define label distribution requirements to support DSDefine how DS packets are to be treated in an MPLS network

Mapping from the DS Code Point (DSCP) to the forwardingbehaviour of an MPLS LSP

Define how DS behaviour aggregates are supported byMPLS networkMapping is to two types of LSP

LSPs which use the EXP bits to specify the class (E-LSP)LSPs where the class is obtained from the label value

Page 37: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 37/54

5. QoS & CoS –DiffServ and MPLS (cont.)

E-LSP The EXP field of the MPLS header is used by the LSR to

determine the class applied to the packet

Mapping DSCP to E-LSP Ingress LSR maps a subset of the 64 possible DS values to one

of 8 possible values, or Behaviour Aggregate (BA), defined bythe EXP (3 bits) field

Page 38: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 38/54

5. QoS & CoS –DiffServ and MPLS (cont.)

DS and MPLS can provide QoS becauseDS provides scheduling and queuing TE provides guaranteed bandwidth

DiffServ providesScalable support for CoS applied to traffic aggregates

MPLS-TE providesOptimised use of resources and selection of routes with BW

constraint

Page 39: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 39/54

6. DS-TE

DS-TEBasicsUsesExample

Page 40: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 40/54

6. DS-TE –Basics

Called DiffServ-Aware MPLS Traffic Engineering or DS-TECombining MPLS-TE and DiffServ

QoS on a per-class basisApply TE on a per-class basis

Separate BW reservation is required for each traffic classRequire admission control on a per-class basis

Page 41: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 41/54

6. DS-TE –Uses

Uses of DS-TE To limit the proportion of classes on a link

E.g. Ensure that VoIP uses no more than a certain percent of the BWto meet delay requirements

To maintain proportional BW allocationE.g. 3 traffic classes (1 [45%], 2 [35%] and 3 [20%])These proportions of link BW to be maintained regardless of

established order, rerouting etc

To provide a guaranteed BW serviceDedicate a DS class to the guaranteed trafficPolice the traffic at the ingress against the traffic contractPlace a limit on percent link BW usedConstraint-based routing used to enforce

Page 42: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 42/54

6. DS-TE –Example

DS-TE example The TE path may have sufficient total BW for all classes but

insufficient BW for a given class The DS-TE path sets up LSPs and allocate BW on a per-class

basis: QoS on a per-class basis

Page 43: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 43/54

7. MPLS vs. FR

MPLS vs. FRGlobal comparison FR label switchMPLS label switchNetwork topologyRoutingNetwork engineeringBandwidth guaranteesBandwidth useCoS

Page 44: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 44/54

7. MPLS vs. FR –Global comparison

RMDCN Site

RMDCN Site

RMDCN Site

RMDCN Site

Local loop rate

Point topoint CIR

RMDCN SiteRMDCN Site

RMDCN SiteRMDCN Site

RMDCN SiteRMDCN Site

RMDCN SiteRMDCN Site

Local loop rate

Point topoint CIR

RMDCN Site

RMDCN Site

RMDCN Site

RMDCN Site

Local loop ratePoint toany CAR

RMDCN SiteRMDCN Site

RMDCN SiteRMDCN Site

RMDCN SiteRMDCN Site

RMDCN SiteRMDCN Site

Local loop ratePoint toany CAR

FR VPN Point-to-point PVCs make up the

VPN A Committed Information Rate (CIR)

per circuit

MPLS VPN Global connectivity to any other sites

belonging to the VPN A Committed Access Rate (CAR) par

site

Page 45: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 45/54

7. MPLS vs. FR –FR Label Switch

Labelling done by CPE, which “speaks”FR

Page 46: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 46/54

7. MPLS vs. FR –MPLS Label Switch

Labelling done by Edge Router: CPE is MPLS-agnostic(classic IP router)

Page 47: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 47/54

7. MPLS vs. FR –Network Topology

FR Network topology is on a customer-

per-customer basis Network path = PVC

MPLS A single shared network topology Network topology designed by

provider

Page 48: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 48/54

7. MPLS vs. FR –Routing

FR Routing is managed CPE-to-CPE Backbone does not participate to

routing

MPLS Routing is managed by the backbone CPE exchange routing information

with the backbone

Page 49: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 49/54

7. MPLS vs. FR –Routing (cont.)

FR When no direct PVC is available,

routing involves several hops

MPLS Optimised routing: directly sent to

destination

Page 50: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 50/54

7. MPLS vs. FR –Network Engineering

FR PVCs’topology: between which CPEs? Which bandwidth for which PVC? Which bandwidth for access lines?

MPLS Which bandwidth for access lines?

… and that’s it! All core network design done by

network provider

What is left for customer to decide?

Page 51: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 51/54

7. MPLS vs. FR –Bandwidth Guarantees

FR In the SLA: site-to-site guarantees Backbone is dimensioned accordingly Guarantees are met due to technique

MPLS In the SLA: guarantees on global traffic

sent by given site to all other sites Shared paths in the backbone Guarantees should statistically be met

Page 52: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 52/54

7. MPLS vs. FR –Bandwidth Use

FR Unused bandwidth of a

PVC can not be transferredto another PVC

MPLS Bandwidth can be spent

anyhow

Page 53: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 53/54

7. MPLS vs. FR –Class of Service

FR Can be achieved by subscribing a

dedicated PVC for each CoS 20 PVCs with no CoS 40 PVCs with

2 CoS (costs!) CPE is responsible for packet

classification (applicationrecognition)

MPLS Built-in: shared paths per CoS on the

backbone CPE is responsible for packet

classification and coloring (DSCP tag) Edge LSR maps DSCP to appropriate

path

Page 54: MPLS Technology Overview

RMDCN workshop on the migration to IP/VPN 54/54

Questions?